← AI security in plain English

AML.TA0011

MITRE ATLAS / Impact

The adversary tries to manipulate, disrupt, erode confidence in, or destroy AI systems and data, using techniques such as Denial of AI Service, Erode AI Model Integrity, and Cost Harvesting.

Think of it likeTrashing the house on the way out: smashing things, jamming the locks, and running up the utility bills out of spite.

In plain English

Attackers cause the real damage: knocking the AI offline, degrading its accuracy, running up huge compute bills, or using its outputs to harm people and erode trust.